摘要：Performing quantitative evaluation of forest fire severity scientifically and reasonably is helpful to revealing the changing of forest ecosystems under fire, and is also of great significance for studying the vegetation recovery and management. Taking the north rim of Grand Canyon National Park in USA as the study area, combined with the composite burn index ( CBI) after field survey, the authors used Landsat5 TM images of Poplar Fire to analyze the applicability of NDVI, NBR,ΔNDVI andΔNBR so as to evaluate fire severity. According to the result obtained, there is some difference between the four remote sensing indices in identifying forest fire intensity of different levels. For non-fire and light fire, indices from a uni-temporal can perform better than indices from bi -temporal (pre and post fire), and NBR has the highest accuracy up to 66. 7% and 80%, respectively; on the contrary, for moderate fire and severe fire, indices from bi-temporal ( pre and post fire) can perform better than indices from a uni -temporal, and ΔNBR outperformed the others, because it considers only indices difference resulting from change of vegetation situation and environmental factors caused by forest fire and not affected by surroundings;it has high accuracy of evaluating moderate fire and severe fire, with the accuracy up to 100% and 90%. In general, indices from bi-temporal ( pre and post fire) have higher overall accuracy than indices from a uni-temporal, and ΔNBR has the highest overall accuracy in evaluating fire severity with the accuracy up to 86. 2%, which is hence the most suitable remote sensing indices to evaluate fire severity in this study area.%科學合理地定量評估林火烈度,對揭示林火干擾下森林生態系統的變化,以及植被的恢復與管理具有重要意義。以美國科羅拉多大峽谷國家公園北緣的Poplar Fire為實驗區,利用Landsat5 TM影像,結合實地調查的綜合火燒指數( composite burn index,CBI),分析評價了歸一化差值植被指數( normalized difference vegetation index ,NDVI),歸一化火燒指數( normalized burn ration, NBR),差分歸一化植被指數( differenced normalized difference vegetation in-dex,ΔNDVI)和差分歸一化火燒指數(differenced normalized burn ration,ΔNBR)4種遙感指數對林火烈度評估的適應性。結果表明,4種遙感指數對識別不同等級林火烈度存在一定的差異。在未過火區和輕度火災區,單一遙感指數的精度略高于差分遙感指數,其中NBR的提取精度最高,分別達到了66.7%和80%；在中度火災區和重度火災區,差分遙感指數的精度高于單一遙感指數,ΔNBR的提取精度最高,分別達到了100%和90%。總體上,基于差分遙感指數的林火烈度制圖精度總體高于單一遙感指數,其中ΔNBR的總體制圖精度最高,達到了86.2%。因此,ΔNBR是林火烈度分析與評估的適宜遙感指數。
